Lab 7: Digital Change Detection

Introduction:

This exercise was intended to teach the class different methods of measuring change in land use/land cover over time.  Different methods were used to qualify and quantify change through visual methods, post-classification change detection, and a model that mapped detailed from-to changes in land use/land cover.


Methods:

Write Function Memory Insertion:

Two images, one from 1991 and one from 2011 of Eau Claire and surrounding counties were compared using write function memory insertion.  First, the red band of the 2011 image was stacked with the near infrared band of the 1991 image.  Then the color guns were set with the red color gun displaying the 2011 red band and the blue and green color guns were set to the 1991 NIR band.  This gave an image that was able to display where most of the change occurred based on brightness and coloring (Figure 1).  The areas that are the brightest red are the areas that have experienced the most change.  This method is able to show where change has occurred but doesn't quantify it in any way.  Also there is no from-to change information available.

It can be seen from this image that most of the change that occurred in the area appears to have been near he rivers or in the urban areas.  (Figure 1)

Post-Classification Comparison Change Detection:

Post-classification comparison change detection provides the from-to information that write insert memory function did not and is much better method at assessing quantitative changes.  Two classified images of the Milwaukee Metropolitan Statistical Area (MSA) for 2001 and 2006 were assessed using this method.

The first step involved looking at the measurements of the different classes to simply quantify percentage changes and put them into a table (Figure 2).  This was simply done by looking at the measurements in the attribute table and doing some simple conversions to get the measurements into hectares.

It was found that the majority of the change occurred in bare soil and open spaces, at least as far as percentage change.  This could be due to more development taking place in these areas as they are much more able to be effected than urban areas.  Also there are less of these areas so the smallest changes have the highest percentage of change.  (Figure 2)

Mapping this change is crucial in many applications, particularly environmental assessment and monitoring.  Of particular interest in this exercise was the change from wetlands to urban, forest to urban, agriculture to urban, wetlands to agriculture, and agriculture to bare soil.  These various changes were mapped out using he Wilson-Lula algorithm (Figure 3).  From here all of the images were brought into ArcMap and a map was generated to show the change that occurred (Figure 4).

The corresponding classes were all loaded into the algorithm at once in model maker in order to produce five different images of each of the desired changes.  (Figure 3)

This is the map generated to show the various types of change that occurred and where they occurred over the period from 2001 to 2006.  (Figure 4)

Conclusion:

Digital change detection is a very useful and applicable part of remote sensing.  Not only is it useful but there are many different methods of performing it.  Different considerations should be made when assessing which type of digital change detection should be run particularly whether one wants a quantifiable result or simply wants to display change location.  Through this lab, the class has become much more comfortable with performing different methods of digital change detection and being able to interpret the results.